The broad, long-term goal of this proposal is to position disease modeling as a viable tool for clinical decision making and policy development. The specific goal is to utilize the three complementary surveillance models developed by the CISNET prostate group to address some of the most persistent and pressing policy questions in prostate cancer. In the proposed work we will extend the models to capture downstream outcomes following diagnosis, such as disease recurrence and secondary treatment. The extensions will be informed by some of the largest and richest population-based data sources available and will be validated extensively using these datasets and results from recently published US and European prostate cancer screening trials. The extended models will be used to project the expected costs and benefits of different screening and treatment policies in order to identify those likely to be of most value in practice. The screening policies will consider different ages to start and stop screening, inter-screen intervals, PSA-based criteria for biopsy referral, and combination policies that incorporate novel screening biomarkers. The treatment policies will include immediate versus delayed primary treatment and immediate versus deferred secondary treatment following biochemical failure. We will extensively investigate the ramifications for both disease-specific and other-cause mortality of policies that include hormonal therapy, the most common systemic treatment for suspected or confirmed metastatic disease. Recognizing that different policies may be called for in different subgroups, we will also investigate the need for targeted policies within subpopulations defined by factors known to affect prostate cancer risk and outcomes, namely age, comorbidity, race, and obesity. The proposed work comprehensively covers the continuum of cancer control issues amenable to modeling that face prostate cancer investigators today. We plan to conduct the research using a coordinated comparative modeling approach in which independent models are standardized and made comparable by the use of common inputs and the implementation of common base case scenarios. .